To further expand services for University of Georgia researchers utilizing AI, The Georgia Advanced Computing Resource Center will deploy innovative, in-demand equipment to its High-Performance Computing cluster this summer.

This $2.4 million investment by the university will support research computing needs for faculty in a variety of disciplines, including artificial intelligence, data science, and bioinformatics, and will be available for use starting in the Fall 2024 semester.

UGA’s Enterprise Information Technology Services (EITS) team will install 26 new Graphical Processing Unit (GPU) compute nodes on the Sapelo2 cluster and three additional compute nodes with significant RAM memory (three terabytes each) to address specific workloads. The GPU nodes are detailed as 12 quad-H100, 2 quad-A100 and 12 quad-L4 NVIDIA GPUs GPU compute nodes are highly specialized processors adept at specific mathematical operations. Originally developed for gaming and video rendering applications, GPUs have become the processor of choice for AI-related applications, with orders of magnitude increases in processing time. 

Guy Cormier, director of research computing for EITS, leads the team responsible for deploying and supporting the new equipment. This team has been exploring recent advances in AI and GPU technologies to support the research community that will use this service. 

“The investment will allow UGA researchers to experiment with AI approaches more rapidly, test new ideas, and refine AI models much faster,” Cormier said. “Overall, the impact of these resources on AI research at UGA will be overwhelmingly positive, accelerating the pace of discovery, enabling researchers to tackle more ambitious problems, and fostering collaborative projects across multiple disciplines.” 

Now that equipment has arrived at the Boyd Data Center on campus, work will begin this summer to install the servers with power and networking, configure the operating system, and embed these compute nodes into the Sapelo2 High-Performance Computing cluster. These compute nodes should be fully functional and ready for use by the start of the Fall 2024 semester.

Investments of this size and complexity require tremendous executive support and collaboration.

“We would like to thank EITS Vice President Tim Chester and Provost S. Jack Hu for their support in providing funding needed to acquire this significant new computational resource for UGA’s research community,” said Dr. Alan Dorsey, associate CIO for research in EITS and professor of physics at UGA.

Once in use, Dorsey noted that this investment will allow UGA “to further expand and accelerate computational services to the AI and data-intensive research community.”

For more information about the Georgia Advanced Computing Resource Center, please visit https://gacrc.uga.edu/or contact Dr. Alan Dorsey at [email protected].